The present invention relates to methods and apparatuses for enhancing performance of optical fiber systems, and networks and systems usable therewith.
Optical fibers, due to their extremely high bandwidth capability and electromagnetic interference (EMI) immunity, have been extensively developed in recent years and are rapidly replacing other types of communication media. Specifically, in recent years extensive research and development into all aspects of optical fiber technology has rapidly moved the technology from the drawing board into the laboratory and into the field in commercial settings.
Distribution optical links differ fundamentally from long haul point-to-point links since in the latter the cost of the entire link, including signal generators, receivers, and repeaters, is distributed over the large number of subscribers served by the multiple signals handled by the link, whereas this is not possible with many distribution links. For example, a distribution optical fiber ring architecture network has been proposed as illustrated in FIG. 1 for servicing a plurality of stations 50 with an optical fiber 51 arranged so as to form a closed loop, with each station including an optical receiver 52 for reading information from the loop, and an optical generator 53 for transmitting the read information or alternatively new information onto the loop. Protocols are used to indicate the destination of information on the loop, and in general many stations 50 read and transmit the same information before it reaches its intended designation. A major disadvantage of such an architecture is that it comprises a plurality of point-to-point connections which requires that the information be regenerated numerous times to get around the loop, and since the loop comprises a plurality of series connections, unique backup systems need to be incorporated to prevent the entire system from shutting down when any one station ceases to properly operate. For an optical point-to-point distribution link, the cost of the dedicated optical generator, receiver, and fiber generally exceeds their electrical counterparts, and hence electrical communication media continue to be preferred over optics in point-to-point distribution applications.
It has long been known that potential enormous cost savings are obtainable if a nonpoint-to-point distribution link or network could be economically produced to replace point-to-point distribution links, and several commercially viable nonpoint-to-point architectural networks have been developed for electrical communication media, such networks generally being referred to as "bus" architectures or networks, one example being disclosed by Biba, U.S. Pat. No. 4,365,331.
As used herein throughout, the terms "bus architecture" and "bus network" comprise any multiple station nonpoint-to-point system wherein information normally passes through the system without having to be regenerated each time it passes a station. However, though nonpoint-to-point distribution networks have been economically produced, the major portion of these networks use electrical wire, rather than optical fiber, as the communication medium, even though optical fiber is recognized to offer numerous and significant advantages over wire as a communication medium.
One of the major reasons optical fiber has not been widely used in place of wire in such networks is that nonpoint-to-point optical fiber distribution networks are not sufficiently cost effective as compared to nonpoint-to-point wire distribution networks and when compared to point-to-point wire distribution networks, despite the extensive effort to develop nonpoint-to-point optical fiber distdribution networks. Rather, even though numerous nonpoint-to-point optical fiber communication networks have been proposed and developed in the prior art, each one suffers one or more serious disadvantages which results in serious cost problems when transferring the technology from the drawing board to the laboratory and from the laboratory to the field in a commercial setting. To date, the common problem to those having all levels of skill in the art has been the lack of development of a network capable of servicing a sufficiently large number of subscribers per repeater spacing to adequately reduce the cost per subscriber of the entire cost of the network.
The number of subscribers capable of being serviced by any given network per repeater can be limited by either (1) available network power and inherent power losses or (2) bandwidth, and as explained in more detail below, these limitations are not independent.
Regarding the first limitation, since optical transmitters, receivers, and fibers have finite operational ranges in that only finite maximum amounts of power can be injected into and supported by an optical fiber, and a finite minimum amount of optical power is required to detect the information, each network has an inherent power dynamic range of operation. For example, if a network is capable of generating and supporting as much as 10 mW of optical power and the network receivers are capable of detecting the information at a minimum optical power of 1 uw at a given bandwidth and given bit error rate, the power dynamic range of the network is 40 dB. For all known prior art bus architectures, each additional tap adds a finite amount of power attenuation to the network, and in this example assuming each tap adds 1.0 dB attenuation, it can easily be determined that the network is capable of servicing a maximum of 40 taps per repeater, the actual number of taps being less than this since some power must be reserved to compensate for losses in the optical fiber, splices connecting optical fibers, etc.
Regarding the second limitation, even if the network has sufficient power to service 40 taps per repeater, if the total bandwidth required by subscribers being served by the taps exceeds the available bandwidth of the network, then the number of subscribers or taps must be reduced or the number of repeaters increased, even though sufficient optical power otherwise exists in the network. For example, if each tap continuously requires 1.0 mHz bandwidth for subscribers serviced thereby, and if a fixed time multiplexing scheme is utilized which is capable of handling +mHz bandwidth, then only 30 taps could be serviced per repeater.
In addition, the first limitation (power) is dependent on the second limitation (bandwidth). Specifically, optical receiver sensitivity varies inversely to the operating network bandwidth, and accordingly if the bandwidth of the network increases from level A to level B, a receiver that is capable of determining information by detecting an optical signal as small as -30 dBm (30 decibels below 1.0 mw) at a given bit error rate at bandwidth A requires an optical signal higher than -30 dBm to achieve the same bit error rate at the higher bandwidth B, this being due to the fact that receiver noise increases directly with the square root of the bandwidth. Accordingly, as the bandwidth of the network is increased to accommodate additional taps and subscribers, the power dynamic range is reduced rendering the number of taps and subscribers the network can accommodate less; hence both limitations one and two set forth above must be mutually considered and satisfied. In addition, in absolute terms, in general any given network can be modified to increase the power dynamic range or bandwidth, but each such modification adds cost to the network and oftentimes produces other disadvantages. For example, the power dynamic range of a network can easily be increased by replacing moderately priced optical receivers with higher priced top-of-the-line performing optical receivers; however, the cost increase incurred thereby may not justify the benefits obtained. Alternatively, a network designed for a monomode optical fiber can be modified to utilize a graded index multimode optical fiber which can support more optical power and hence increase the power dynamic range of the network, and even if the cost differential between the types of fibers is not significant, this modification will reduce the network bandwidth since monomode fiber is capable of transmitting information at higher bandwidths than graded index fiber. Hence, a variety of considerations require analysis to devise any kind of suitable optical fiber distribution network.
Numerous efforts have been directed to improving optical fiber bandwidth, and a major disadvantage of these efforts is that many of the solutions attenuate the optical power of the network an undue amount and/or add undue cost.
A significant limitation on bandwidth is due to modal dispersion, e.g., the tendency of different light modes of a single optical signal to propagate at different group velocities axially down an optical fiber. Modal dispersion results in pulse spreading which for step and graded index multimode fibers is a severe limitation on available bandwidth. Numerous approaches have been taken in the prior art to minimize or eliminate the effects of modal dispersion and hence increase the bandwidth of optical fibers and of networks incorporating these approaches. Common approaches have been to use various types of mode scramblers, mode strippers, or mode filters which eliminate outermost or slowest modes being supported by the fiber, the removal being accomplished by downward coupling of the outermost modes into lower order modes, or simply eliminating the outermost modes. Examples of such approaches are discussed by Marcatili, U.S. Pat. No. 3,777,149; Kaiser, U.S. Pat. No. 3,969,016; Gloge, U.S. Pat. No. 3,785,718; Midwinter, U.S. Pat. No. 3,944,811; Midwinter, G.B. 197,808; Dyott, U.K. specification 1,420,458; Storozum, "Mode Scrambling Can Enhance Fiber-Optic System Performance", Electronics, Feb. 24, 1981, pages 163-66, see page 166; Sakaguchi, Japanese Kokai 55-29847 (A); and Yanase, Japanese Kokai 52-32341.
Numerous other approaches have been taken, examples of which are Tien, U.S. Pat. No. 3,617,109; Jackson, U.S. Pat. No. 4,125,768; Cohen, U.S. Pat. No. 4,447,124; Marcuse, U.S. Pat. No. 3,909,110; Eve, U.S. Pat. No. 4,205,900; Ueno, Japanese Kokai 52-49040; and Bennett, "Extending the Range of Long Wave Length Multimode Optical Fibre Transmission Using Decision Feedback", SESSION A XII:SYSTEM (I). However, all these approaches are either unduly complex, absorb excessive optical power, or present other difficulties.
Though enhanced bandwidth is an important consideration in developing an optical fiber distribution network, in particular a bus network, a more fundamental problem has been how to access or "tap" the optical fiber in a passive manner so as to be able to inject information onto the network (e.g. "write") and/or withdraw information from the network (e.g. "read") such that light representative of the information remains in the network in amounts sufficient to allow a sufficiently large number of stations to be connected to the network per repeater to make it economically attractive. Numerous efforts have been directed by the prior art to developing suitable taps for nonpoint-to-point distribution networks, e.g. bus networks, and these efforts have met with very limited commercial acceptance due to complexities of components embodied therewithin (e.g. high cost) and operational limitations posed by such embodiments which result in relatively few stations being able to access the network per repeater.
For example, Love, U.S. Pat. No. 4,072,399, discloses a distribution network useable for either a ring or bus architecture which utilizes a plurality of taps 17-22. However, the network, and in particular the taps, are intricate in construction and craft sensitive to install, and a further disadvantage is that each tap produces an excessively large excess loss. The term "excess loss" as used herein represents the fraction of power normally expressed in per cent or dB, which is attenuated by a tap but not actually detected thereby, or passed on by the tap.
Polczynski, U.S. Pat. No. 4,089,584, describes an optical fiber bus architecture network wherein a plurality of subscribers sequentially inject light into a first portion of an optical fiber and subsequently sequentially withdraw light out of another adjacent section of the optical fiber 14. This network requires the use of an optical fiber cable which includes a core 14 having at least one planar side, with prism-type couplers 20 being disposed in contact with the planar side of the core 14 by removing a portion of a cladding 12 in a vicinity of each coupler. Accordingly, this network is disadvantageous in that it requires a specialized form of fiber, e.g. a fiber having a core with one rectangular side, as opposed to an ordinary fiber having a circular core, requires that a portion of the cladding of the optical fiber be removed prior to tapping which inherently is a time intensive and expensive procedure, and requires that a prism be disposed adjacent an exposed portion of the core where the cladding has been removed. Accordingly, the component parts making up the network are expensive as well as the time required to assemble it, and the network has a further disadvantage in that removing the fiber cladding is a factory operation that results in needing one splice on each side of the tap and hence a high excess loss.
Biard, U.S. Pat. No. 4,400,054, discloses a bus architecture network wherein a plurality of subscribers are radially connected with a rectangular scrambler rod 20 via eight rectangular optical fibers 31-38, which in turn are connected to further optical fiber arms 31a-38a via prisms 31b-38b. Again, the unique shape of the various optical waveguides disclosed by Biard, and the relatively complicated interconnected structure created thereby, render this network nonadaptable for widespread use due to cost.
Singh, U.S. Pat. No. 4,234,969, is another example of an optical fiber bus architecture which utilizes an intrically constructed and hence relatively expensive optical tap 18. The taps, illustrated in FIG. 2 of the reference, incorporate multiple reflecting surfaces therewithin which result in relatively high excess losses.
Palmer, U.S. Pat. No. 4,317,614, discloses a bus architecture utilizing optical taps 18, 24, 34, etc. each of which comprises first and second bent fibers 126, 128 which have had confronting surfaces lapped or ground and subsequently interconnected to create optical coupling therebetween. Lapping optical fiber surfaces as disclosed is an extremely craft sensitive procedure and hence expensive, and again the excess losses resulting when the fiber cladding is substantially removed are unduly large due to splice losses.
Ozeki, European patent application publication 0,080,829, discloses several unique architectural network designs aimed at increasing a number of subscribers served by the network. However, though the designs each have the advantage of increasing a number of subscribers otherwise capable of being served by the network, each design has a common deficiency in that the couplers required are unduly complex and expensive, and furthermore the number of subscribers capable of being served by each network design is relatively low in view of the relatively high excess losses of the taps used in the network.
Steensma, U.S. Pat. No. 4,450,554, discloses a bus architecture utilizing a star coupler, and a common disadvantage of networks utilizing star couplers, as is well known, is that the excess losses imposed by star couplers is relatively large, being on the order of 2 dB and the power is divided by the number of subscribes, which necessarily limits the number of subscribers which can be served by the network.
It has long been known that light can be withdrawn from an optical fiber cladding at a bend, as taught by Kapany, N.S., "Fiber Optics; Principles and Applications", Academic Press, San Francisco (1967), that light withdrawal from the cladding is facilitated by using an optical coupler in contact with the fiber, as taught by Fujimura, U.S. Pat. No. 3,801,389, Goell et al., U.S. Pat. No. 3,982,123 and Miller, U.S. Pat. No. 3,931,518, and that light can be injected into an optical fiber at a bend using an optical coupler, as taught by Maslowski, German Offenlegungsschrift 2,064,503 (FIG. 4). A disadvantage of light taps which operate on optical fiber claddings at bends is that the stresses generated at the bends oftentimes tend to fracture the optical fiber, and this problem is aggravated since removal of the buffer to optically couple with the cladding exposes the cladding to humidity which quickly and seriously degrades the strength of the fiber and its ability to remain intact when bent and stressed.
Campbell et al., European Publication Number 0,063,954, discloses a method and means for injecting light into and withdrawing light from an optical fiber at a bend without removing the buffer so as to allow temporary in situ local launch and detect techniques to be used for aligning optical fibers prior to splicing. The method and means comprises disposing a bent portion of an optical fiber against an optical coupler such that injected light passes through the optical coupler and the fiber buffer as it is injected into the fiber core, and light to be withdrawn passes through the fiber buffer and into the optical coupler as the light is withdrawn from the fiber core. Since the buffered optical fibers are only bent temporarily during the aligning and splicing operations, the probabilities of fiber breakage are minimized.
To date, preferred taps for distribution network applications comprise star couplers, reflective, and biconic couplers. Each of these couplers can be provided as a separate component, with star couplers being capable of splitting an optical signal into as many as 128 smaller signals, an excess loss of such star couplers being in a range of 1-3 dB. Referring to FIG. 2, a star coupler 80 comprises a plurality of optical fibers 81, fused together at a central region 82 such that an optical signal 84 propagating into the central region from any one of the fibers on one side of the central region is split by the central region into a plurality of smaller signals 85 and propagate down each of the fibers extending from the other side of the central region, as illustrated by arrows in FIG. 2. The 1-3 dB excess loss created by such a star coupler is represented by arrow 83.
A biconic coupler 86, illustrated in FIG. 3, is similar to a star coupler, and comprises two optical fibers fused together at a signal splitting central region 87, the central region generally being formed so as to split an incoming large signal 88 into a relatively small drop signal 90, generally equal to 1-10% of the intensity of the signal 88, and a relatively large residual signal 89. Biconic couplers have excess losses 91 in excess of 0.5 dB.